1

In W. McGregor (ed.), Expression of possession, 213-248. Berlin & New York: Mouton de Gruyter.

Plain vs. situated possession in Czech: a constructional account

[preproofs version]

Mirjam Fried, Czech Academy of Sciences, Prague

1. Introduction* The focus of this paper is the kind of possession that can be identified

broadly as a time-stable relation that is presupposed (not asserted) between two

entities, a possessor (PR) and a possessum (PM). It is well known that this

relation can be expressed in a number of ways even in a single language, let alone

cross-linguistically; what still remains to be worked out in sufficient detail is the

exact nature of the variation and the relationships among the variants. My purpose

here is (i) to identify the factors that help differentiate between syntactically

distinct expressions of this kind of possession and (ii) to propose a way of

representing the patterns in a network of grammatical constructions organized

around their shared features. The illustrative material comes from authentic Czech

usage, both written and spoken, as attested in the Czech National Corpus.1

Heine (1997: 143) labels the presupposed, time-stable possession as

“attributive possession”, so categorized in contrast to “predicative possession”,

and this categorization includes characteristic differences in syntax as a crucial

criterion: phrasal syntax for the former, clausal for the latter. This distinction,

though, is too general to give us a realistic picture of the variety of possessive

expressions. The non-clausal expressions tend to come in several distinct formal

variants which represent a rather diverse set, both in their form and their function.

I will examine a subset of such variants in Czech and offer a functionally and

cognitively oriented analysis that is motivated by three general questions: (i)

What kind of linguistic knowledge is necessary for native-like production and

comprehension of these possessive patterns? (ii) How can the speakers‟

knowledge and understanding be adequately represented? And (iii) can the

representation help us make more precise claims about attributive possession as a

linguistic category, with implications beyond accounting for the Czech facts.

Some attributive possessive expressions are exemplified in (1-3); all of

these patterns are familiar and found in many languages. The examples in (1) are

two variants of an adnominal PR in a NP headed by the PM.2 (1a) is an agreeing

1 The material used here comes from roughly 400 million-word written corpora (SYN2000,

SYN 2006PUB), supplemented by a sample from Czech language and literature (LITERA,

SYNEK), and a 2.5 million-word spoken corpora (PMK, BMK, ORAL2006). 2 The distribution of these two forms is conditioned morphologically: the post-nominal

genitive is obligatory if the PR is morphologically neuter (1b), or a multiword NP, or a plural

noun, or is modified by a relative clause. In other cases, the agreeing pre-nominal form is the

neutral choice.

2

modification structure, in which the PR is marked by possessive morphology and

agrees with the noun in case, gender, and number, like any pre-nominal modifier

in Czech. In (1b), the PR is in the genitive and obligatorily follows the head noun.

Conceptually, both of them could be treated as cases of “possessor specification”

(Heine 1997:167), but I will refer to them collectively as Genitive PR (GP),

choosing a more traditional and familiar label. In (2), both the PM and the PR are

syntactically independent NPs and the PR status is associated with the dative

marking on the PR. This pattern is known as “external possessor” (e.g. Payne &

Barshi 1999) and is typically associated with PR affectedness; I will, therefore,

call it Affected PR (AP). The example in (3) shows a strategy in which the PR is

not explicitly identified.3 For easier orientation, the possessive expressions will be

in bold throughout the paper.4

(1) Genitive PRs (GP)

a. Pre-nominal

Natali-in-y rodiče ty maj furt

Natalie.PR.F-NOM.PL parents.NOM they have.PRES.3PL always5

{ňákej státní svátek , tak sou doma pořád .}

„Natalie‟s parents, they have {state holidays all the time, so they‟re always

at home}‟ [PMK137;148304]

b. Post-nominal

Ohneme nožičky dítěte v kolenou

bend.PRES.1PL leg.ACC.F.PL child.GEN.SG.N in knee.LOC.PL

„We‟ll bend the baby‟s legs at [the] knees‟ [SYN2000;34388742]

(2) Affected PR (AP)

spravil nám i auta [SYNEK:1598227]

fix.PST.SG.M 1.PL.DAT also car.ACC.PL

„he even fixed our cars‟

(3) Implicit PR:

ten pán ztratil brejle {a nemohl je najít}

that man.NOM.SG.M lose.PST.SG.M glasses.ACC

„that man lost [his] glasses {and he couldn‟t find them}‟ [LITERA;1680536]

3 There are additional variants, including a morphological adjective as an expression of PR

status. I will not be concerned with these marginal cases here but the general approach

provides ways for incorporating such patterns as well. 4 A note on presenting the data: when the surrounding text is necessary for clearer

understanding, it will be enclosed in curly brackets {} and left without interlinear glossing. 5 Abbreviations in glosses: PR „possessor‟, NOM „nominative‟, GEN „genitive‟, ACC

„accusative‟, DAT „dative‟, INS „instrumental‟, LOC „locative‟, M/F/N „masculine/feminine/

neuter‟, SG/PL „singular/plural‟, SPRL „superlative‟, RF „reflexive‟, PRES „present tense‟,

PST „past tense‟, IMP „imperative‟, INF „infinitive‟, NEG „negative‟, AUX „auxiliary‟, AP

„affected PR‟, GP „genitive PR‟.

3

My primary concern will be the relationship between (1) and (2);

reference to the Implicit PR variant is necessary mainly as a background against

which the explicit patterns can be studied. The example in (3) is simple: even the

minimal context given here is enough to suggest that the owner of the lost glasses

is the subject referent, rather than some other person. The preferred interpretation

has to do with what we know about glasses as common personal possessions.

Other times the implicitness allows a greater range of possible interpretations,

especially if the sentence contains more than one potential candidate for the PR

status, including direct discourse participants, or if the PM is something other than

a body part. Speakers know to infer the appropriate configuration based on

conventional expectations about possible possessive relationships vis-à-vis

particular context. However, if it is the case that a possessive relation can be (and

in Czech very often is) left implicit, then we have to ask what reasons speakers

might have for choosing the explicit options and particularly, what – if anything –

conditions the choice of (1) vs. (2).

I will be concerned only with the second (easier) question here, pursuing

the hypothesis that the differences should revolve primarily around the

(in)alienability of the PM and the affectedness of the PR as two independent,

competing factors; the hypothesis is motivated by the patterning found in many

other languages. However, the analysis will demonstrate that the interaction

between affectedness and (in)alienability is systematically more complex than

this: specifically, in cases of conflict, affectedness takes precedence over

inalienability in licensing the AP form. This leads, among other things, toward an

absolute prohibition on PM (of any kind) as a transitive subject and almost

equally strong prohibition on subjects of active intransitive verbs. Given that

similar syntactic restrictions on the grammatical role of the PM are cross-

linguistically common in external possessor constructions (Payne & Barshi 1999),

the present analysis highlights a semantic explanation that may apply well beyond

the Czech facts. Finally, mapping out the interaction between affectedness and

(in)alienability also provides a generally applicable approach to incorporating

peripheral cases in which a non-possessive expression may acquire a possessive

construal.

In order to keep the analysis reasonably focused, I will take as my starting

point a relatively narrow definition of the concept of possession. I exclude the

broad sense of belonging in part-whole relations in which the whole is not an

animate entity (Heine‟s 1997: 35 “inanimate possession”), but my conception is

not quite as narrow as Taylor‟s (1989: 202-203) notion of prototypical ownership.

Like Taylor, I will take possession to be an “experiential gestalt”, which

presupposes that the PR is a sentient being and has specific reference. However, I

will consider a broader range of possessa than his prototype allows (“specific

concrete things”). Finally, I will also work with the notion of possessibility

hierarchy. There seems to be no firm consensus about the exact shape of a

universally applicable hierarchy; the main points of variation seem to be the

relative placement of kin relations, inherent attributes, and clothing, and their

4

status evidently depends on particular languages. For the purposes of this study, I

find it sufficient to assume a very general scale along the lines of Payne & Barshi

1999: body parts > kin > close alienable > distant alienable. However, a close

analysis of the Czech AP (section 3) will suggest a more refined scale, partially

overlapping with Tsunoda‟s (1995: 576) Possession Cline.

By examining the range of factors that motivate the distribution of the GP

and AP patterns, I will argue that the distinction goes far beyond reducing the

issue to treating the two strategies as two formal variants of a conceptual schema

called “possessive specification” (Heine 1997: 167), each motivated by a

particular instantiation of an existential event schema – Goal vs. Genitive (p. 47).6

Both of these variants merely assert the existence of a presupposed possessive

relation, each in a different syntactic form. The Genitive schema is associated

with the pattern [PR‟s PM]), while the Goal schema, here corresponding to the

AP pattern, takes the form [PM (exists) for/to PR]. Both schemas are said to

correlate with permanent inalienable possession. Neither of these broad schemas

thus helps explain the fact that the Czech GP and AP have very different

distributions in actual discourse and that the difference in form between (1) and

(2), as well as some special syntactic constraints associated with (2) but not with

(1), follow from a number of co-occurring properties: the inherent meaning of the

PM and the PR, the involvement of the PR in the event expressed by the clause,

verb semantics, and information flow. I will also suggest that the presence or

absence of some of these properties correlates with particular speech situations.

The two forms will thus be best treated as grammatical constructions in

the sense of Construction Grammar (esp. Fillmore 1989, Croft 2001, Fried &

Östman 2004), and I will also draw on the notion “cognitive frame” (Fillmore

1982) for incorporating the possessive relation in the constructional

representations. I will argue that each possessive variant constitutes a

semantically and pragmatically distinct pattern – a conventional cluster of

semantic, pragmatic, and morpho-syntactic properties – and that each pattern

represents a functional prototype within a network of possibilities for expressing

attributive possession.

The paper is organized as follows. The semantic and pragmatic properties

of the PR and the PM are discussed in sections 2 and 3, respectively, addressing

the issue of a possessive prototype. Section 4 studies the interaction between the

possessive relation and the structure and meaning of the sentence in which it

occurs. All of this comes together in section 5, in which I present the patterns (1)

and (2) as two grammatical constructions that occupy overlapping domains within

the functional space of attributive possession. I illustrate the ways in which the

constructions in the network may interact both with each other and with the

possessive prototype, suggesting a more systematic account of the ways in which

6 Only these two schemas in Heine‟s typology are relevant here. Unlike, for example, Russian

or Romanian, Czech does not have any locative-type patterns ([PM (is) at PR) expressing

possession.

5

the prototype can be extended to more peripheral instances. Section 6 briefly

concludes the paper.

2. Semantic and pragmatic features of the PR

The inherent semantics of a quintessential PR does not predict, by itself, anything

about the linguistic form and behavior of the noun phrase that encodes the PR; the

referents in (1-3) are all specific human beings. Moreover, both AP and GP are

sometimes possible for a given possessive relation, as shown in (9); note,

however, that they are not quite synonymous.

(4) a. Před pěti lety nám zemřel otec. [SYNEK;2279741]

before 5 years 1PL.DAT die.PST.SG.M father.NOM.SG.M

{Matka se zhroutila. Z nemocnice ji propustili, ale není v pořádku.}

„Five years ago, father died. {Mother fell apart. She‟s back from the

hospital, but she‟s not in good shape.}‟

b. Před pěti lety zemřel náš otec.

before 5 years die.PST.SG.M 1PL.PR.NOM.SG.M father.NOM.SG.M

„Five years ago, our father died.‟

Other times, though, only one form is available. For example, the AP in (5a)

cannot be replaced by GP (5b) and the examples in (6) demonstrate the reverse:

(5) a. V uších nám hvízdal vítr [SYN2000;2062266]

in ear.LOC.PL 1PL.DAT whistle.PST.SG.M wind.NOM.SG.M

„Wind was whistling in our ears.‟

b. * V našich uších hvízdal vítr

in 1PL.PR.LOC.PL ear.LOC.PL whistle.PST.SG.M wind.NOM.SG.M

(6) a. Z nějakého […] důvodu ho její vlasy

from some.GEN reason.GEN 3SG.M.ACC 3SG.F.PR.NOM.PL hair.NOM.PL

nesmírně vzrušovaly. [SYNEK; 941353]

immensely excite.PST.PL

„For some [unknown] reason, her hair excited him enormously.‟

b. *… vlasy jí ho nesmírně vzrušovaly

hair.NOM.PL 3SG.F.DAT 3SG.M.ACC immensely excite.PST.PL

Similar effects have been noted for various languages (e.g. O‟Connor 1994, Croft

1985, Berman 1982, Chappell & McGregor 1995, Manoliu-Manea 1995, Payne &

Barshi 1999), including Czech (e.g. Zimek 1960, Pit‟ha 1992, Fried 1999a, 2008),

and the subsequent discussion will show that the contrast has to do both with the

involvement of the PR in the reported event and the relative communicative

prominence of the PR and PM in a given piece of discourse.

2.1 Affectedness of PR

6

The relationship exemplified in (5) illustrates the fact that inalienable

possession overwhelmingly prefers AP in Czech, to the point of leaving it as the

only expressive option in some cases (i.e., if the speaker has a reason to mention

the PR explicitly in the first place, instead of opting for Implicit PR). The case in

(6), then, appears to be an „exception‟ since the only possibility there is GP, even

though the PM is a body part like in (5).7 However, this apparent contradiction is

fully motivated and I will revisit it in sections 4 and 5. First let us consider a less

categorical situation: examples where AP appears to be replaceable by a GP form.

The expressions in (a) are corpus attestations of AP, the (b) examples are their GP

counterparts (constructed).

(7) a. {Pak přišla průvodčí} a proštípla nám lístky

and punch.PST.SG.F 1PL.DAT ticket.ACC.PL

{a cesta doplynula ve vší pohodě} [SYNEK;2261787]

„{Then the conductor came} and punched our tickets {and the rest of the

trip went all smoothly}.

b. proštípla naše lístky

punch.PST.SG.F 1PL.PR.ACC.PL ticket.ACC.PL

(8) a. “vsaďte se, že se tomu ťulpas-ovi policajtsk-ému vysmolím

bet.IMP RF that RF that.DAT nitwit-DAT cop.ADJ-DAT poop.PRES.1SG

před dveře na schody {a ještě si s ním přitom budu povídat !}”

in.front.of door onto staircase [SYN2000;87610]

„ “you can bet that I‟ll take a dump on the steps in front of the nitwit cop‟s

door {and keep chatting with him in the process, too!}” ‟

b. …že se vysmolím před dveře toho ťulpas-a policajtsk-ého

that.GEN nitwit-GEN cop.ADJ-GEN

(9) a. Když jim vzali peníze měnovou reformou,

when 3PL.DAT take.PST.PL money.ACC currency.ADJ.INS.SG reform.INS.SG

{prodávali koberce, šperky a obrazy, jen aby mohla zůstat doma.}

„When [the government] took their money in the currency reform, {they

kept selling rugs, jewelry and paintings, anything [to make it] possible for

her to stay at home.} [SYNEK;653077]

b. Když vzali jejich peníze

when take.PST.PL 3PL.PR money.ACC

The issue in (7-9) is not the incompatibility between the PM and the GP

form, as is the case in (5). The problem in (7-9) is the degree to which the PR is

involved in the depicted events. The example in (7) is the most flexible one: GP

(7b) implies that the conductor punched some tickets that belonged to us but were

7 Whatever observations are made in this paper about the behavior of body parts as the PM,

they apply to all specific candidates of body part status. Czech does not make any grammar -

coded distinctions between different types of body parts.

7

not necessarily related to our riding the train. The AP (7a), in contrast, presents

the tickets as necessary for the ride, we were holding them in our hands, handed

them over, and then took them back from the conductor, to hang onto them. The

GP form is in principle possible but somewhat odd, given what we know about

the usual ways in which passengers and conductors interact. A GP form in (8)

would be even more problematic since its implication would be contradicted by

the subsequent coordinated clause: GP (8b) would be felicitous whether the cop is

present or not when the speaker is squatting in front of the cop‟s door; the

automatic interpretation actually would be that the cop is not around. Yet, the

point of the speaker‟s bet is that he will be simultaneously having a conversation

with the cop. The contextual incompatibility is still stronger in (9): GP (9b) would

be possible whether or not the PR was alive during the currency reform, while AP

(9a) is felicitous only if the PR was alive. The AP emphasizes this state of affairs,

which makes it the only coherent choice for the follow-up about the consequences

the PR suffered (selling off property to make ends meet).

To summarize, the two forms display a systematic division of labor that

can be related to Bally‟s 1995/1926 notion of personal domain or (subjectively

applied) indivisibility. GP expresses plain possession in the broadest sense, where

the concept of indivisibility plays no role. In contrast, AP casts the possession

relation as something that is relevant to the PR in a particular way, as something

in his sphere of interest beyond just the fact of being owned. AP signals that the

PR is being affected (positively or negatively) by something that affects the PM

(cf. also Chappell & McGregor 1995). In this light, it is not surprising that certain

types of possessa strongly prefer AP, in contrast to GP: the tighter the possessive

relationship, the greater the chance that manipulating the PM will directly affect

the PR. Body parts are the extreme on the continuum since they are truly

inseparable from the PR, and hence the unacceptability of (5b). The GP form can

only be interpreted in one way: the ears are not attached to the PR‟s body (are not,

that is, bona fide body parts) but are some ear-like objects, physically detached

from any bodies.

This extreme restriction applies only to the pre-nominal GP, though; as

shown in (1b), the post-nominal genitive is sometimes attested with body parts

and the form maintains the true body part reading. We thus have to conclude that

the relative flexibility between allowing both AP and GP extends all the way to

the top of the possessibility hierarchy. Nevertheless, a difference in form (AP vs.

GP) always correlates with a shift in interpretation, no matter what the PM. For

example, the choice of GP with a body part in (1b) is motivated by the type of

discourse and the focus of attention. This token is taken from instructions about

how to use a rectal thermometer to take a baby‟s temperature. The

communicatively relevant issue here is how to manipulate the baby‟s position so

that the person succeeds in what needs to be done; the author of the instructions is

not concerned with what (dis)comfort it may bring to the baby, which would be

the only possible reading if an AP form were chosen. A similar analysis applies to

example (4), in which a kinship relation is cast in a different light depending on

8

whether AP or GP is used. The GP in (4b) would imply that the speaker was

estranged from, or at least not very close to, his father and therefore unaffected by

his death, in contrast to (4a), which unequivocally implies that the father‟s death

had tangible consequences, as is clear from the subsequent text.

Two generalizations emerge from these facts: (i) inherent (in)alienability

is not a good predictor of AP vs. GP encoding (as also noted explicitly by Bally

1926/2995 and Chappell & McGregor 1995) and (ii) the PR in the AP form plays

a special role in the event expressed by the predicate, while no special status is

associated with the GP variant.

2.2 Information structure

As a syntactically independent NP, the dative PR is free with respect to its

position in the clause relative to the PM. It can, therefore, participate in

articulating topic-focus relations within the possessive relation itself, and thereby

accommodate the nuances of structuring information flow, which in Czech is

expressed through word order (Firbas 1966, Daneš 1974, Sgall et al. 1986, Grepl

& Karlík 1998). On the other hand, the PR position within GP is restricted: the

pre-nominal PR cannot be very easily separated from the PM (except under very

special contextual circumstances) and the genitive is obligatorily fixed in the post-

nominal position. Consequently, GP is dispreferred or outright impossible in

certain contexts simply for reasons of information structure, even if it might be

appropriate on semantic and syntactic grounds. Consider (10):

(10) a. {Za něco se styděl,} bál se podívat lidem do tváře,

fear.PST.SG.M RF look.INF people.DAT in face.GEN.SG.F

{když jim něco nabízel} [LITERA; 1947648, otapavel]

„{He felt embarrassed about something,} he was afraid to look people in

their FACES {when he was selling them something}‟

b. bál se podívat do tváří lidí,

fear.PST.SG.M RF look.INF in face.GEN.SG.PL people.GEN

„…he was afraid to look in the faces of PEOPLE‟

The example in (10a) gives discourse prominence to the faces, as indicated

by the small caps in the English translation. The point of the sentence is to

describe the subject‟s reluctance to face people. Replacing AP by a GP expression

(10b) would shift the focal status onto the genitive PR, simply by virtue of its

phrase-final position, and this reconfiguration would sound very odd in the

context of this passage. (10b) creates the expectation that lidí „of people‟ is to be

understood as contrastive focus: either in contrast to the faces of other entities

mentioned earlier, or as the head of a restrictive relative clause that would have to

follow („he was afraid to look in the faces of [those] people who he was selling

something to, but not other folks‟). Such readings are clearly unintended here, and

9

using AP is a general strategy for making the PR and the PM available for

structuring information flow independently of each other.

The infelicitous character of (10b) cannot be easily attributed to the

disembodiment reading associated with pre-nominal GP, for two reasons. First,

the GP pattern as such does not imply disembodiment, as we see in (5a); in

that sentence, the PR‟s hair is unambiguously attached to her head. Moreover,

since the post-nominal GP is obligatory with certain grammatical types of PRs,

listed in footnote 1 (here the issue is the plural of the PR noun), the inherent

semantics of the PM becomes irrelevant. And second, since the information

structure value of the PR can be freely manipulated in the AP, it is possible to

form a discourse-structure equivalent of (10b) as well, shown in (10c) below:

(10) c. bál se podívat do tváře lidem

fear.PST.SG.M RF look.INF in face.GEN.SG.F people.DAT

„…he was afraid to look in the faces of PEOPLE‟

This AP form is subject to exactly the same interpretive restrictions as (10b),

namely putting the people into a focus role, whether contrastive or plain. The

only difference is that the post-nominal GP cannot be preposed, under any

circumstances.

To summarize, AP gives the PR a certain prominence that GP does not

provide. However, this kind of prominence must be understood as an event-based

prominence (i.e., treating the PR as an affected event participant, as opposed to

just an owner), independent of a discourse-based prominence (i.e., assigning

information structure roles within the possessive relationship), which can be

manipulated separately. I will revisit the implications of this feature in section 4.

3. Semantics of the possessum

As illustrated in (11), the head noun in the GP pattern can be semantically

anything: concrete, abstract, animate, inanimate.

(11) Han-in-y ruce / sestry / kamarádky / knihy / názory / povinnosti

Hana-POSS.F-NOM.PL hands / sisters / friends / books / opinions / duties

This semantic freedom suggests that the GP pattern simply marks a very general

relationship between two entities, which can be conceptualized as denoting a unit

of sorts. This conceptualization accommodates possession (including inalienable)

in the narrow, experiential sense as well, but the form evidently is not restricted to

encoding truly possessive relations. The semantic and functional breadth is

corroborated by the fact that the same form is found not only in combinations

such as (11), all expressing (at least loosely understood) possession, but also in

various common „genitive‟ functions (not discussed here) or in the „syntactic‟

10

function, i.e. encoding verbal arguments in nominalized expressions. The latter is

exemplified in (12), where the pre-nominal possessive form Janovu „Jan‟s‟ marks

the agent role of the noun pomoc „help‟, not any PR of help.

(12) děkoval mi za Jan-ov-u pomoc [LITERA; 3101771]

thank.PST.SG.M 1SG.DAT for Jan-POSS.M-ACC.SG.F help.ACC.SG.F

„he thanked me for Jan‟s help‟

The AP pattern is clearly different from GP in that AP does not cover the

syntactic function shown in (12); we cannot create an AP paraphrase of (12).

However, paraphrasing the expressions in (11) is possible, at least in principle.

Thus, considering the AP examples in (7–9) above and (13) below, AP might not

appear to present a dramatically different picture from GP with respect to the

kinds of possessa it permits, along the full possessibility hierarchy:

(13) a. Komolil jména lidem při nejrůznějších příležitostech

distort.PST.SG.M name.ACC.PL people.DAT at SPRL.various occasion.LOC.PL

„He mangled people‟s names on all kinds of occasions‟

[LITERA; 375950]

b. když jsme byly tomu strejčkovi na tom pohřbu

when AUX.1PL be.PST.PL that.DAT uncle.DAT on that.LOC funeral.LOC

„when we were at that uncle‟s funeral‟ [PMK441:786749]

c. jak se nám rozpadá společný stát

as RF 1PL.DAT fall.apart.PRES.3SG common.NOM.SG.M state.NOM.SG.M

„as our shared country is splitting [into two]‟ [SYNEK; 1953603]

It has been noted in the external PR research that external PRs tend to co-occur

with possessa at the high end of the possessibility hierarchy, with different cut-off

points in different languages (Payne & Barshi 1999, Payne 1997a, b). We have

now seen that the Czech AP roughly follows the same scale of preferences but it

is worth checking this general tendency against the semantic range attested in

actual discourse. The sample in (2), (7-9), and (13) can be easily extended by

additional possessa found with AP in CNK: pytel „sack‟, hračky „toys, vodovod

„water-pipe‟, svět „world, cesta „journey‟, sebevědomí „self-confidence‟, smlouva

„contract‟, život „life‟, svatba „wedding‟, reforma „reform‟, králík „rabbit‟, kůň

„horse‟, etc. We can see that even this fairly randomly assembled list covers the

whole spectrum of the possessibility hierarchy and contains both concrete and

abstract entities. At the same time, it must be acknowledged that abstract nouns

generally do have limited currency in the AP pattern; it is impossible to either find

or construct coherent examples with nouns such as hodnota „value‟, krása

„beauty‟, kvalita „quality‟, chyba „mistake‟, moudrost „wisdom‟, výmluva

„excuse‟, nadřazenost „superiority‟, while any of these would be perfectly

compatible with the GP pattern.

11

On a closer look, however, the possessa that are found in the AP pattern

fall into several conceptual categories, which can be properly identified in relation

to conventional cultural expectations, rather than in terms of general linguistic

categories such as alienability, concrete/abstract, animate/inanimate, etc. The

corpus material suggests a number of salient semantic classes of items possessible

either individually or collectively by human beings and construable as such within

the speakers‟ cultural experience. The list in (14) does not necessarily represent a

real hierarchy in Czech but it is of course possible to trace certain parallels with

the linguistically defined possessibility hierarchy that is more explicitly attested in

other languages (cf. Tsunoda 1995):8

(14) a. things that are part or features of self (body parts; name, title; speech; life;

doubt, memory, intention, self-confidence, right to decide, etc.)

b. members of ‘family’, understood broadly as a culturally established unit of

shared domestic life (kinship relations; pets and other domestic animals)

c. garments and their parts

d. environment perceived as essential to our existence, including dwellings

and their parts (world; house, door, plumbing, bathroom; prison cell;

backyard)

e. objects useful in an individual’s daily life (cars, toys, flashlight, money,

tickets, guitar strings)

f. common activities and established rituals (journey, wedding, funeral,

graduation, education, vacation, holiday)

g. social and/or political organization (state/country, constitution, reform)

This list makes it evident that there is no blanket prohibition on the usage

of abstract nouns. Moreover, the attested combinations cannot be simply replaced

with GP without changing their meaning, as we have seen in section 2. It is

important to note, though, that the abstract concepts that co-occur with AP tend to

come from particular semantic domains and are experientially based: they have to

do with mental or physical states and cognitive capacities typical of human beings

(14a) or with personal and social rituals (14f). The latter can be easily extended

further into more specialized contexts, in which a possessive construal can apply

to abstract concepts that are inherent in various types of social institutions in

general, such as political organization and public life (14g), and where the PR is,

therefore, a collective, not an individual. Granted, the categories in (14f-g) do not

represent the same sense of ownership/possession as the classes in (14a-e) and as

such are somewhat removed from the central, prototypical definition of

possession. Nonetheless, they are not the only source of abstract nouns in the AP

8 This group can also be conceptually further decomposed into more specific categories, some

of which are also suggested by Tsunoda (1995): parts of body, representations of self,

cognitive and speech qualities, manifestations of self-awareness, etc. Such subcategorization,

however, leaves the main point of the generalization in (14a) for Czech unaffected, since

these additional conceptual distinctions seem to have no grammatical reflexes.

12

pattern and so they do not invalidate the general observation about the

possessibility of abstract entities.

On the other hand, we also find a number of both abstract and, especially,

concrete entities that do not seem to fit easily into any of the categories in (14) at

all: branka „goal/score [in soccer]‟, brankář „goalie‟, pacient „[hospital] patient‟,

tramvaj „street car‟, dřevo „wood‟, smlouva „contract‟, etc. Taken out of context,

these items seem entirely random and might suggest, yet again, that the AP

pattern is not that different from GP in the range of possessa it permits. However,

„context‟ is the operative word here; these items are always invoked and

interpreted within specialized contexts (e.g., medical care, transportation,

commerce), sometimes even in distinct types of discourse (sports reporting),

which frame possessive relations in ways specific to those contexts and establish

the conditions that allow the PR to be cast as affected in particular ways. No such

contextual framing is required for the use of the GP variants. And it is also worth

pointing out that these context-sensitive possessive relations are, therefore,

different from the possessa in (14f-g): those two categories are relatively

independent of special contexts and can be understood as being inherent (and in

some ways universal) parts of common human experience. By contrast, the

context-dependent possessa are more transient and non-essential.

We thus must conclude that the distribution of GP and AP in actual usage

is subject to various preferences and sometimes even inviolable constraints with

respect to the type of PM. Relevant generalizations can be formulated as follows:

(i) body parts cannot occur in the pre-nominal GP pattern and, instead, require

AP; (ii) the range of possessa that can naturally co-occur with AP is semantically

more restricted than with GP; (iii) the restrictions can be meaningfully articulated

only in terms of culturally based clusters of concepts, sometimes in combination

with specific types of discourse contexts, not in terms of purely linguistic

distinctions, as has been traditionally accepted; and (iv) the distribution of GP is

more restricted in terms of information structure.

4. Constraints on verb semantics and clause structure

4.1 Verb semantics

The semantic and pragmatic differences between AP and GP outlined in

the preceding sections have different consequences for the types of predicates

each form co-occurs with. Predictably, the distribution of GP is entirely

independent of any verb meaning, since GP expresses pure possession. In

contrast, the dative PR‟s involvement in the event depicted by the verb makes AP

sensitive to the semantic type of the verb. The affectedness of the PR, which is

tied to the affectedness of the PM, presupposes events that are semantically

compatible with affectedness. This is, indeed, a correlation well attested with the

AP pattern. AP strongly prefers „contact‟ predicates, such as spravit „fix‟ (2),

13

proštípnout „punch through‟ (7a), čistit „clean‟, postřílet „shoot down‟, narovnat

„straighten‟, roztrhnout „tear‟, přivázat „tie to st.‟, hodit „toss‟, umýt „wash‟, as

well as other affective verbs, such as komolit „distort‟ (13a), ničit „ruin‟, osvěžit

„refresh‟, etc.

The corpus material shows that another very well-represented class

concerns verbs of removing, such as vzít „take away‟ in (9), sežrat „gobble up‟,

vyloučit „expel‟, krást „steal‟, ztratit „lose‟; these verbs presuppose a possessive

relation and are thus particularly good candidates for accommodating affected

PRs, by expressing possession-removal situations. Among intransitives, verbs of

states or spontaneous processes without any identifiable instigator are highly

compatible with AP semantics and richly attested in the corpus as well: zemřít

„die‟ (4), rozpadat se „fall apart‟ (13b), zřítit se „collapse‟, rozbít se „break down‟,

padat „fall‟, zmizet „disappear‟, mrznout „freeze‟, zešílet „go crazy‟, cukat

„twitch‟, třást se „shiver‟, smrdět „stink‟, etc.

Predicates that do not fit these semantic profiles are very hard to come by.

The following, showing the syntactically transitive verb vidět „see‟, is only an

apparent counterexample to this generalization:

(16) {není to proto, že ti drbani jsou nadáni mimořádnou jasnozřivostí a}

vidí lidem do kapes [LITERA; 727505; frybort2]

see.PRES.3PL people.DAT into pocket.GEN.PL.F

„{[if a businessman gets robbed...] it‟s not because these lowlifes are

endowed with extraordinary clairvoyance and} see into people‟s pockets‟

First of all, the use of AP in (16) can be explained by appealing to information

structure, just like we saw in (10): the use of the post-nominal GP kapes lidí

„pockets of people‟ (the only GP possibility here due to the plural of the PR noun)

would shift the focus away from the pockets, which would make the form

contextually problematic. The context clearly centers on what happens to the

owners of the pockets; the passage gives explanations about the circumstances

under which people get robbed. But equally important is the verb itself: it is not

used here in its purely perceptual meaning of „having a visual experience‟, which

would require an accusative-marked perceptum. Instead, the directional phrase do

kapes „into the pockets‟ indicates a more active reading along the lines of „look

inside‟. The sense of pure perception is completely incompatible with AP, as

shown in (17); this combination is incomprehensible and is equally impossible

even with body parts as PM:

(17) *vidí lidem kapsy

see.PRES.3PL people.DAT pocket.ACC.PL.F

„(s)he sees people‟s pockets‟

The corpus data thus confirm the conclusions of previous studies, namely,

that non-affective verbs generally require a GP form. The following examples

14

further illustrate the prohibition on APs with such verbs, in contrast to the attested

GP, both with transitive (18) and active intransitive (19) predicates:

(18) a. ted’ znám eště její rodiče [PMK143; 519767]

now know.PRES.1SG also 1SG.F.PR parents.ACC

„now I know her parents too‟

b. *ted’ jí znám eště rodiče

now 3SG.F.DAT know.PRES.1SG also parents.ACC

(19) a. {N.P. měla dokonce kurzy v jazykovce zadarmo ,}

když ta její máma tam pracovala .

when that.NOM.SG.F 1SG.F.PR mom.NOM.SG.F there work.PST.SG.F

„{N.P. could even take courses in the language school for free} since her

mom was working there‟ [oral2006; 175302]

b. *když jí tam ta máma pracovala .

when 3SG.DAT.F there that.NOM.SG.F mom.NOM.SG.F work.PST.SG.F

The impossibility of the (b) forms cannot be explained by the semantics of

the PM; recall that kinship relations preferentially select AP (4a, 13b). Instead, it

must be attributed to the AP‟s semantic incompatibility with predicates that do not

allow the PR‟s affected role in the depicted event to be properly integrated with

the meaning of the predicate. GP, on the other hand, does not place any such

requirement on the verb meaning and becomes, therefore, the only available

alternative. Notice also that the affectedness must be facilitated by the predicate

semantics, not just a potentiality motivated by the inferences suggested by the

broad context. This difference is illustrated in (16) and (19) above. In (16), the

relevant verb meaning is signaled overtly by the morphology of the second

argument. The case in (19), with the intransitive verb pracovat „work‟, is more

subtle. An AP variant (19b) cannot be substituted even though the general context

clearly supports the inference that the PR (i.e., the daughter, identified as N.P.) is

positively affected by the fact that her mother works in a particular place; N.P.‟s

benefit is the central concern of this utterance. I will revisit this problem in

section 5, to consider potential counterexamples.

The co-occurrence patterns attested with AP vs. GP thus lead to two

generalizations: GP is fully independent of verb semantics, while AP can only

occur with predicates that are semantically compatible with affectedness. The

latter can be formulated in terms of a hierarchy of preferences as in (20), with a

clear cut-off point (indicated by the double arrow >>) between active intransitive

and weakly transitive predicates; the latter are fully excluded from the AP pattern:

(20) {strongly transitive, non-active intransitive} > active intransitive

>> weakly transitive (Vs of perception and cognitition)

15

4.2 Syntactic constraints on the possessum

These verb-related semantic preferences alone cannot account for certain

additional facts about the AP pattern; we still need to examine the cases

exemplified by (6), here repeated as (21), which involve a particular syntactic

constraint on the possessum. In (21), the main predicate, vzrušovat „cause

excitement‟, is affective and the PR is a pre-nominal GP, which should not be

possible with a body part PM; recall the discussion of (5). Yet, (21a) is the only

grammatical way of expressing the possessive relation in this sentence and cannot

be replaced by an AP under any circumstances (21b):

(21) a. Z nějakého […] důvodu ho její vlasy

from some.GEN reason.GEN 3SG.M.ACC 3SG.F.PR.NOM hair.NOM.PL

nesmírně vzrušovaly. [SYNEK; 941353]

immensely excite.PST.PL

„For some [unknown] reason, her hair excited him enormously.‟

b. *… vlasy jí ho nesmírně vzrušovaly

hair.NOM.PL 3SG.F.DAT 3SG.M.ACC immensely excite.PST.PL

Since GP is inside a self-contained NP, we can expect its distribution to be

quite free across all syntactic slots. The AP, on the other hand, follows the cross-

linguistically commonly attested pattern of excluding the PM of an external PR

from certain syntactic functions. In Czech, the PM can be a transitive object (2,

7a, 9a, 13a), an oblique complement (5a, 8a, 10a, 16), and an intransitive subject

(4a, 13c). However, we see in (21b) that AP cannot appear as a transitive subject,

even if it is a body part. The same prohibition is evidenced in (1a) with a kinship

PM, here repeated as (22a): GP is the only way to express the possessive relation

here, while AP (22b) is ungrammatical. The same restriction extends to the

subjects of active intransitive verbs, as we saw in (19) above.

(22) a. Natali-in-y rodiče ty maj furt

Natalie.PR.F-NOM.PL parents they have.PRES.3PL all.the.time

ňákej státní svátek , {tak sou doma pořád . }

some state holiday.ACC.SG.M

„Natalie‟s parents, they have some kinda state holidays all the time, so

they‟re always home‟ [PMK137;148304]

b. *Natali-i maj rodiče furt ňákej státní svátek

Natalie-DAT have.PRES.3PL parents.NOM …

In order to motivate the syntactic restrictions on the PM, we have to appeal again

to the affectedness requirement. The PM cannot appear in event roles that

presuppose agentive participants.

The conclusion we can draw from the syntactic behavior and the

combinability with different predicate types is the following. There are two

16

competing factors for choosing the AP vs. the GP strategy – PR affectedness and

(in)alienability of the PM. Generally, AP preferentially attracts inalienable

possessa and for these, it is the only expressive possibility vis-à-vis pre-nominal

GP forms. However, the (verb-determined) affectedness of the PR evidently takes

precedence over inalienability, leading toward absolute prohibition on possessa

(of any kind) as transitive subjects and almost equally strong prohibition on

subjects of active intransitive verbs.

4.3 Phrasal syntax and the attributive possession

We have already noted the readily obvious syntactic difference between

GP and AP: GP is so called because the PR is a modifier syntactically

dependent on the head noun denoting the PM. The GP is thus a particular

(semantic) variant of a more general Modification construction, which is a

cluster of syntactic, morphological, and semantic properties defining a

particular type of attributive relationship between a noun and its modifier,

following a simple phrasal template: [ModP-or NP-um]NP for the pre-nominal

form and [NP-um ModP-or,GEN]NP for the post-nominal GP.

In contrast, it is hard to see on what definition of constituency we could

argue that the Czech dative-marked PR forms a single syntactic constituent

with the PM. The example in (23) makes it clear that the Czech AP is not an

adnominal structure. The corpus example in (23a) shows GP used in an

adposition to a noun phrase and (23b) demonstrates that the dative form is

prohibited; (23b) is severely ungrammatical:

(23) a. {Ve švédském Stokholmu žije od dětství, kdy tam}

jeho maminka, sestra Jiřiny, emigrovala.

his mom.NOM.SG.F sister.NOM.SG.F Jiřina.GEN.SG.F emigrate.PST.SG.F

„{[He]‟s been living in Stockholm, Sweden, since childhood,} when his

mom, Jiřina‟s sister, emigrated.‟ [SYNEK; 11289608; 1997]

b. *jeho maminka, sestra Jiřině, emigrovala.

his mom.NOM.SG.F sister.NOM.SG.F Jiřina.DAT.SG.F emigrate.PST.SG.F

The PR and the PM have to be analyzed as two autonomous NPs, both of them

incorporated as ordinary complements into the structure and the meaning of a

sentence. In fact, both the form (dative) and the semantics of affectedness place

AP in the family of other dative complements.9 These include „thematic‟ datives

9 In terms of grammatical roles, the dative complements could be classified as indirect

objects, with the understanding that a two-place predicate in Czech can encode its arguments

using the pattern [subject – indirect object], e.g. pomoci „help‟ in (24). Since it is more

informative to refer to these complements either through their case form (always dative) or

their semantic role status, I choose not to label them in terms of grammatical roles here. The

point is that Czech dative NPs are primarily motivated semantically, not syntactically in terms

of grammatical roles (cf. also Fried 1994).

17

(required by the predicate as one of its arguments, namely one that is in some

non-manipulative, perhaps „mental‟ way affected by the event in which it

participates) or adjuncts that are added to verbs that do not require a dative-

marked argument but can accommodate an extra participant that can be construed

as a beneficiary or a maleficiary of the expressed event. A thematic dative is

exemplified in (24) with the verb pomoci „help‟, showing also the fact that not all

of these datives are restricted to animate referents. An adjunct dative is in (25),

with the verb otevřít „open‟, and for these datives, the referent is necessarily

human; the datives are in bold.

(24) a. pomohl mnoha lidem a dětem

help.PST.SG.M many.DAT.PL people.DAT and child.DAT.PL

„he helped lots of people and children‟ [SYNEK, 10733898;1997]

b. trend, kterému pomohl především

trend.NOM.SG.M which.DAT.SG.M help.PST.SG.M especially

{silný příliv zahraničního kapitálu} [SYNEK; 8751696; 1999]

„a trend which was helped particularly {by a strong flow of foreign

investment}

(25) Tento balón otevřel lidstvu oblohu.

this balloon.NOM.SG.M open.PST.SG.M mankind.DAT.SG sky.ACC.SG.F

„This balloon opened up the sky for mankind.‟ [SYN2000; 137864]

This connection has been explored in detail elsewhere (Fried 1999a, b)

and I will not revisit it here beyond summarizing that the AP shares the special

affectedness with other dative-marked roles (experiencer, recipient, beneficiary/

maleficiary) and particularly with the usage shown in (25), which I label Dative of

Interest (DI). Both DI and AP require the referent to be human and in both cases,

the dative NP is an extra element not required by the valence of the verb.

However, the datives are incorporated into the sentence as if they were full-

fledged arguments. But AP adds a particular feature that is absent in the other

dative roles, including DI: the presupposed possessive relation between the

referent of the dative and something else in the sentence. This is an important

point that bears on the issue of framing possessive relations in terms of broad

cognitive schemas, such as the Genitive vs. Goal-based schemas suggested in

Heine‟s (1997) typology. It is of course true that at some very abstract level, all of

the datives, including AP, relate to the concept of goal-ness. But it is also very

saliently the case that AP is distinct from other goal-expressing roles and that the

distinction is not just a matter of interpretation but has systematic reflexes in

semantic constraints and syntactic behavior. In the following section, I will

suggest a way of re-organizing the conceptual space of attributive possession so

that we can capture more accurately the relationships – commonalities as well as

contrasts – between the relevant semantic categories.

18

5. Czech GP and AP as distinct functional patterns

5.1 Plain possession vs. situated possession

The syntactic behavior only confirms that GP and AP cannot be taken

simply as two alternative and fully comparable expressions of attributive

possession. To capture the essence of the distinction, we can label the patterns as

Plain Possession and Situated Possession, respectively. What they both share,

necessarily, is a human PR and a pre-existing possessive relation. We could

formalize this common background in the form of an interpretive frame,

POSSESSION, roughly outlined in (26). The frame represents a conventionally

established knowledge structure that schematizes the speakers‟ understanding of

prototypical ownership as a particular relation between two entities (PR and PM),

each of which can be associated with various properties that are motivated by

both individual and generally cultural experience. Minimally, the prototypical PR

is schematized as a human being that is highly individuated and referentially

specific, and the PM as fitting somewhere along the possessibility hierarchy (FE

stands for „frame element‟):

(26) frame POSSESSION: FE#1 Possessor [+human]

FE#2 Possessum [on possessibility hierarchy]

For our purposes here, it is not crucial to dwell on all the additional details

of this general frame (such as, perhaps, listing a set of preferred properties for

each frame element). I will simply take (26) as a minimal way of representing

speaker‟s conceptual understanding of prototypical possession, which then is

shaped into different instantiations by elaborating on the specific characteristics of

the PR and/or the PM. Crucially, though, this shaping requires reference to both

form and meaning, not just one or the other. The conceptual prototype organized

in the background frame does not, by itself, say anything about the morphological

or syntactic requirements associated with the morphosyntactic strategies for

expressing the possessive relation. Those involve additional layers of constraints

and I will argue that the best way to capture the nature of those expressive

strategies is to treat each pattern as a grammatical construction in the sense of

Construction Grammar, i.e. as a conventionally expected association between the

elements of this frame and their linguistic expression.

5.2 Constructional organization of AP and GPproperties

Let us start by summarizing the properties of the two frame participants

when used in the AP pattern, which is clearly the more constrained of the two

forms; the items in the list below can be read as being in contrast with the

corresponding features of the GP pattern:

19

(27) Situated Possession (AP):

Possessor

- is a participant in the depicted event (i.e., sensitive to V semantics)

- is (indirectly) affected by the event

- is distinct from other datives (semantically, syntactically)

- can fully participate in information structure relations

Possessum

- comes from a semantically defined and restricted class of items

- must be affected by the depicted event

- is prohibited in certain syntactic functions

- can fully participate in information structure relations

Much of the information summarized in (27) can be attributed to the affectedness

requirement. For example, the semantics of the PM is predicted to include entities

that can be manipulated and that are inherently relevant to human beings and the

routines of their daily existence. It also follows that AP attracts particularly

strongly items high on the possessibility hierarchy. At the same time, what we

know about possessive relations, dative-marked affectedness, or the possessibility

hierarchy does not lead to the prediction that any competition between the

preference for inalienable possessa vis-à-vis affectedness must be resolved, as a

rule, in favor of maintaining a consistent affectedness status (hence the

prohibition on possessa as transitive subjects and intransitive agentive subjects).

Nor does it account for two additional features: the fact that both the PR and the

PM can participate in information structure relations and that the dative form does

not cast the PR in a role simply identical with the role of other dative nominals.

There are thus properties of the AP pattern that have to be captured in

some other way, not just as simple consequences of the PR affectedness. Which

brings us to positing AP and GP as complex conventionalized clusters of specific

syntactic, morphological, semantic, and pragmatic properties, i.e., as distinct

grammatical constructions. Using the Construction Grammar formalism of Fried

& Östman (2004), we can represent what speakers have to know about AP as in

Figure 1. The inner box represents the head verb, indicating that AP is dependent

on certain features of the verb: the verb is specified as coming from a semantic

class broadly characterizable as „affective‟ and as having at least one argument

(labeled FE #2) that will be expressed in syntax (as the PM); the latter is indicated

by the val(ence) statement. The rest of the information (i.e. everything in the

outside, larger box) represents properties that are idiosyncratic to the AP pattern,

and these have to do with integrating the possessive relation with the semantic

and syntactic structure contributed by the head verb.

First off, it is the AP construction as a whole, not the head predicate, that

supplies the link to the possessive relation, through the statement [frame

POSSESSION]), which brings along the PR and the PM and all the background

knowledge associated with the representation in (26). Second, the constructional

val(ence) statement in the outer box captures the fact that the syntactic and

20

semantic properties of the argument supplied by the head verb are constrained by

the requirement that its event role be semantically non-agentive (thus ensuring the

prohibition on the PM as a transitive or active intransitive subject). The AP

construction does not specify anything about the PM‟s semantic or syntactic role

(both of these are determined by the semantics of the head verb). However, the

co-indexing (#2) between the head verb, the constructional valence, and the

POSSESSION frame indicates that whatever this argument is in terms of its semantic

and syntactic role in the sentence, it will be interpreted as the PM. Third, the

construction also gives the PR an independent information-structure status

relative to the PM; this must be specified as the construction‟s prag(matic)

property. It follows from this feature that both the PR and the PM are subject to

articulating regular information structure relations, which operate independently

of this particular construction (cf. the discussion in section 2.2); put differently,

both the PR and the PM can appear as either a topical or a focal element,

independently of each other.

The rest of the representation contains features that are shared across AP

and DI, as is indicated by the inherit statement at the top; all the features that

come from this relationship are printed in gray, to show that these specifications

are not unique to the AP construction (strictly speaking, the inherit statement

would be sufficient and all the remaining gray-colored information need not be

spelled out). These features include the following. First, the construction is

syntactically a verb-based pattern (the syn(tax) statement at the top). Second, the

construction itself has additional valence requirements, namely, the PR (#1) is in

the dative (case DAT) and is interpreted as other datives of interest (expressed

through the rel(ation) statement specifying the semantic role, labeled θ). And

finally, it is the construction as a whole that carries the overall meaning of

Situated Possession, spelled out in the sem(antics) statement, as a combination of

the inherited DI semantics and the POSSESSION frame.

21

inherit Dative-of-Interest

sem

#2 [ rel [ 'non-agentive']]}

frame POSSESSION

FE #1 Possessor [human]

FE #2 Possessum [ ]

val

prag [ 'greater discourse independence of Possessor vis-à-vis Possessum' ]

Affected Possessor

syn [ cat v, lex + ]

val

syn [ cat v, lex + ]

sem frame [affective]

FE #2[ ]

{...#2[ ]...}

{ #1 [ rel [ 'interest' ]] ,

[ syn [ case DAT ]]

[ 'circumstances described by the predicate have significant consequences

for the interested party (#1), whose referent is not in control of the event' ]

Figure 1. Affected PR construction

When we try to represent GP as a construction in its own right, its

differences from AP come into sharp relief. To save space, I will focus on the pre-

nominal variant only. Figure 2 shows that the GP construction is a subtype of

general Modification construction (through an inheritance link). Again, all the

inherited information is in gray and is only included for clearer exposition: the GP

construction is a phrasal structure of the category NP (indicated by the syn(tax)

statement at the top of the constructional box) and consists of two constituents

(the inner boxes), with the modifier preceding the head. The information specific

to the GP construction concerns two properties. One is the mapping between the

POSSESSION frame participants onto the two structural daughters, through the co-

indexing: the PR is the modifier, the PM is the head. The other feature is the

cat(egory) of the modifier; I have abbreviated this by simply giving a list of

morphological classes (enclosed in curly brackets): the filler will be either a

possessive pronoun or the special nominal form derived by the possessive suffixes

–ův and –in. Nothing more needs to be specified.

22

role modification

cat [ poss {pro; -UV / -IN }] cat n

syn [cat n]

Genitive Possessor

role head

case

number

gender

#i [ ]

#j [ ]

#k [ ]

morphol. case

number

gender

#i [ ]#j [ ]

#k [ ]

morphol.

prag ['restrict reference of the noun (#2) by property exrpessed in #1']

#1 #2

sem frame POSSESSION

FE #1 Possessor [human]

FE #2 Possessum [ ]

inherit Modification

Figure 2. Genitive PR construction (pre-nominal).

These representations express a hypothesis about the kind of knowledge

speakers of Czech must possess in order to produce and interpret a variety of

concrete linguistic expressions of attributive possession. One part of the

hypothesis is the generalization that the speakers‟ native-like understanding of

such expressions involves the understanding of a rather intricate interplay

between several layers of information, as shown in the figures. But the

constructional analysis and representation can enhance also our insight into the

way these grammatical patterns may be organized in larger networks of distinct

but partially overlapping patterns. Such an organization, in turn, should allow us

to be more precise about the circumstances under which the possessive prototype

can be extended in various directions. I will suggest such a network and its

implications in the next section.

5.3 Constructional network

If we take the concept of attributive possession as a type of functional

space that can be occupied by various expressions of this general possessive

relation, we can organize all the features we have identified as relevant (semantic

as well as grammatical) in a network that shows precisely which features are

shared across individual patterns and which are specific to each pattern. This is

what we see in Figure 3. The shaded area in the middle represents the frame

POSSESSION, with its two crucial participants and the minimal constraints on their

referents listed under each participant (human PR and the relevance of the

possessibility hierarchy). This general possessive relation can be expressed as

Plain Posession, in the upper part of the plane, or as Situated Possession, in the

lower half; we could think of both as more specialized sub-frames, each of which

23

is conventionally associated with a particular form – the GP and AP grammatical

constructions, respectively.

The solid-line rectangle in the upper part of the diagram represents the GP

construction and its relationship to the background frame: the PR must be in the

genitive form(s) and for the PM, we have to note that inalienable possessa are

generally incompatible with, or at least strongly dispreferred in, this construction

(the symbols „<‟, „>‟ indicate the direction of preference). The dashed-line

rectangle represents the AP construction, which modulates the background frame

by adding a number of features to both the PR and the PM, listed in the columns

under each frame participant. With respect to the PM, the most important change

consists in delimiting the semantic classes of permissible possessa (in the diagram

indicated by reference to the classes indentified in (14) and in noting that the AP

construction preferentially selects items from the inalienable end of the

possessibility hierarchy, in contrast to GP.

The dotted-line, rounded rectangles are included to indicate that both GP

and AP constructions overlap in specific ways with other, non-possessive

constructions. In the case of the Syntactic Genitive, the common feature with GP

is just the Genitive form; the construction does not include any part of the

possessive frame and also differs from GP in that the genitive is a participant in

the event denoted by the head noun (agent or patient in nominalizations). AP, on

the other hand, shares all four characteristics of the PR (animacy, event

participant status, affectedness, and dative marking) with the DI construction,

which excludes the possessive semantics of the dative nominal and the AP‟s

constraint on verb semantics. Instead, DI is related to other non-possessive dative

constructions (the space labeled Affected datives). I will return in a moment to the

significance of the three items that are underlined.10

10

Additional constructions expressing attributive possessive relations (such as adjectival,

coming-into-possession patterns based on transfer, etc.) can be incorporated into this network,

once their features are properly worked out.

24

[ Possessor, Possessum ]

human

event partic.

affected

DAT inalienable > alienable

possessibility hiearchy

GEN/poss. morphol. inalienable < alienable

'salient in daily routine' (14)

frame POSSESSION:

Situatedpossession (AP)

Plain possession (GP)

Dat. of interest

Synt. genitive

event partic.

(ex. 12)

(ex. 25)

affected

(ex. 24)Affected datives

'affective' verbs

Figure 3. (Partial) „attributive possessive‟ network.

The network thus captures important facts about the way constructions

may interact and the ways in which linguistic expressions may stretch the

properties of those constructions. I will now briefly comment on three such

examples, all involving the AP construction.

As already noted, a systematic account of possession always faces the

question of what should count as possession and how inclusive or non-inclusive

we should be in defining the PR and the PM. Even with the relatively permissive

prototype assumed in this paper (compared to Taylor‟s), we still have to account

for examples such as (28), in which the PM does not easily fit the categories

suggested for AP in (14):

(28) Ale jednou se nám ztratil jeden pacient

but once RF 1PL.DAT get.lost.PST.SG.M one.NOM.SG.M patient.NOM.SG.M

„But once one of our patients disappeared‟[SYNEK; 1498008,Hrabal 1993]

It is of course hard to argue that there is an ownership relation between the

hospital personnel (here the speaker) and the patients. As discussed in section 3,

the speaker‟s choice to employ AP will be motivated by the discourse context

and/or genre: one that sets up the plausibility of an affective possessive reading

which involves the relationship between a human being and another entity that is

in a contextually salient relation to it. Such cases, then, receive the reading of

situated possession by virtue of being used in the AP construction, and we are not

forced to relax the possessive prototype itself to account for them. Since the

criteria for a plausible PM are evaluated on a sliding scale to begin with, there is

perhaps always room for stretching the scale in a motivated way.

25

A more dramatic departure from the possessive prototype is shown in (29),

where an inanimate entity seems to be cast in a PR-like relation to another

inanimate entity. Since the PR is defined as human (or at least animate), admitting

(29) as a case of possession is a much more serious matter.

(29) {Pokud by půda poklesla jen o padesát centrimetrů,}

uhnijí všem stromům kořeny.

rot.away.PRES.3PL all.DAT tree.DAT. PL.M root.NOM.PL.M

„{Should the soil sink even by as little as 50 cm,} the trees will lose all

their roots to rot‟ (lit. „the roots of all the trees will rot away on them‟)

[syn2006publ; Respect 26/1993]

We could declare that (29) is not a case of possession but a simple part-whole

relation. The advantage of such an analysis would be that it would preserve the

concept of possession as an experiential gestalt, specific to human beings. This

way we would also avoid the danger inherent in any prototype-based analysis: in

order to account for every new deviation, we could, in principle, keep relaxing the

prototype ad infinitum, which then amounts to justifying just about anything as an

instance of the same concept, and the prototype loses its coherence as a tool of

systematic analysis.

However, a categorical exclusion of examples such as (29) leaves

unanswered an obvious similarity between them and the AP pattern, both formally

(the „whole‟ being in the dative) and in the overall affective interpretation: when

the roots die, the tree is certain to die as well. In order to account for these and

similar extensions, the constructional approach offers an alternative that allows us

to incorporate the full range of deviations that may arise in actual discourse, while

at the same time preserving the possessive prototype as the conventional semantic

basis. The AP construction expects the prototype – represented in the frame – to

hold, but of course the match between the prototype and the lexical fillers of the

constructional slots will not be always perfect, stretching the prototype to varying

degrees. In (29), the stretching concerns the semantics of the PR, but at the same

time, it is very close to the prototype in two ways: (i) the PM is (construable as)

inalienable in the same way body parts are and (ii) the mutual relationship

between the whole and its part is fully compatible with the affective meaning of

the AP construction: the whole is affected because its constitutive part is affected.

It is also important to note that GP, shown in (30), does not evoke the situated

possession reading but stays purely at the level of a part-whole relationship. In

(30), the implication may very well be that the trees will somehow make it

anyhow; in any case, (30) is not concerned with the fate of the trees, it is about the

fate of the roots only.

(30) … uhnijí kořeny všech stromů

rot.away.PRES.3PL root.NOM.PL.M all.GEN tree.GEN. PL.M

„...the roots of all the trees will rot away‟

26

I would argue that this difference is due precisely to the fact that the GP

construction is not associated with a special meaning; the form can cover both

possessive and non-possessive relations, and consequently cannot impose a

possessive reading so easily on combinations that deviate from the prototype in a

radical way (such as presenting an inanimate entity as an owner of anything). In

contrast, by invoking the semantic and pragmatic properties of the AP

construction as a whole (especially the affectedness of the PR) we have a

principled way of explaining what allows the stretch into domains in which we do

not have real PRs but only a very specific (and tight) part-whole relationship. In

other words, it is the use of the AP construction in encoding a (close) part-whole

relation that allows a personification reading, i.e. a conceptualization which

mimics a relationship between a PR and a body part.

We could say that both of these cases (28-29) illustrate scenarios in which

the AP construction, as a conventional grammatical pattern in its own right,

facilitates various manipulations of the possessive frame, sometimes also in an

interaction with the closely related, but more general, PART-WHOLE frame11

. But the

AP construction may also attract other patterns in the network and pull them into

an AP reading because of certain shared constructional properties. Here I have in

mind the issue of intransitive agents and their potential for compatibility with an

AP interpretation, in an apparent contradiction to the participant hierarchy

established in section 4.2. It might seem that cases such as (31), with the active

intransitive verb utéct „run away‟, are simply evidence that Czech does not, after

all, restrict the possessa to non-agentive referents and instead excludes only

agents/subjects of transitive verbs.

(31) vašemu tat’kovi utek … ten křeček taky ,

your.DAT dad.DAT run.away.PST.SG.M that.NOM.SG.M hamster.NOM.SG.M also

{to byl první jeho křeček . von byl strašně z toho smutnej . a tak sme mu

museli koupit hned dalšího} [oral2006; 1195644]

„your dad‟s hamster also ran away [in our backyard], {that was his first

hamster and he was all so sad about it and so we had to buy a new one for

him right away}‟

Such an analysis, though, would be an oversimplification since this

patterning, attested very rarely to begin with, is also highly restricted. First, the

utterance has to be available for an affective construal, as is the case here;

replacing the dative in (31) with GP would be incoherent in the light of the

speaker‟s subsequent elaboration about what effect the event had on the PR. And

as noted in section 4.2, this availability has to be accommodated by the verb

meaning. The fact is that only very few verbs appear to work semantically: recall

11

I have not provided the details of this frame or its place in the network, mostly for reasons

of keeping the representation uncluttered and easily readable, given the focus of thi s paper

(pure possession). But it is obvious that it must be part of a more complete representation of

this functional space, particularly in working out the Genitive-related domain.

27

that the quintessential active verb pracovat „work‟ resists an AP reading (19). A

better fit seems to come with verbs of removal (utéct „run away‟, schovat se

„hide‟), which is not surprising, given the semantics of the AP construction as

affecting pre-existing possession. Also verbs of eating or drinking are possible

candidates, presumably because the result of such acts has the potential of

entering the sphere of interest of people other than the eaters/drinkers.12

And

second, the PM (necessarily animate) must be very high on the possessibility

hierarchy. Examples such as (32) cannot be interpreted possessively, even though

there are two animate participants and the action of one (the police) clearly is

intended to have consequences for the other (ostensibly mice, figuratively for the

inhabitants):

(32) {… obklopila naši vilu vozidla Státní bezpečnosti a postupovali tak, }

aby jim neutekla ani myš..

so.that 3PL.DAT NEG.run.away.PST.SG.F not.even mouse.NOM.SG.F

„{…cars of the Secret Police surrounded our house and they [=secret

agents] proceeded [carefully]} so that not even a mouse [could] escape

from them‟ (lit. „to them‟) [syn2006pub; 11696551, LN 33/1992]

We have to conclude that examples such as (31) are more plausibly

analyzed as instances of the DI construction, such as exemplified in (25) and (32),

in which any kind of verb, including all semantic types of intransitives, can be

used, and which expresses a situation with an indirect effect on the „interests‟ of

an animate entity. DI of course overlaps with the AP construction to a great

extent; outside of not having the possessive dimension, the DI only differs in that

it places absolutely no constraints on the verb semantics. It is not a stretch, then,

for a DI token to invite an AP reading, provided that certain features of that token

coincide with a particular narrow set of features of the AP construction.

This finally brings us to the significance of the three items that are

underlined in Figure 3. The features human on the PR and inalienability on the

PM are central to the notion of experientially defined possession. We can,

therefore, expect, that if an inherently non-possessive expression (such as DI)

invites a possessive reading, it can be only at the level of the core possessive

properties. But satisfying these two features of the possessive frame does not, by

itself, guarantee a successful AP interpretation. The PR must also be construable

as affected by the event expressed by the verb, which is a central property of the

AP construction. This is, of course, related to the lexical meaning of the verb,

which is constrained in AP, but that restriction is evidently not as rigid as the

12

While I have not run across such an example in the corpus, I can think of usages such as

(i). The verb semantics found in these combinations is still waiting for more careful research.

(i) Syn se jim opíjel.

son.NOM.SG.M RF 3PL.DAT drink.to.excess.PST.SG.M

„Their son drank heavily [to their worry/embarrassment].‟

28

requirement that, whatever the verb, the PR must come out as an affected entity.

Some verbs with active semantics are inherently better equipped for such a stretch

(e.g. verbs of removal) than others (e.g. pracovat „work‟), and that is what

accounts for the relative (un)availability of DI tokens for an AP interpretation.

The important point is that all these shifts, whether they involve

extensions of the possessive prototype into broader semantic domains or, on the

contrary, attracting tokens of non-possessive constructions, can all be explained

by appealing to the same cluster of properties (the core features of the possessive

prototype associated with AP) and to the AP pattern as a constructional gestalt:

the situated possessive meaning of the whole construction is the „glue‟ that holds

all these seemingly disparate uses together. We could thus think of the three

underlined items as having a privileged status within this network: they constitute

the set of features that are instrumental in various partial shifts, giving rise to and

at the same time constraining novel usages.

6. Conclusions

Perhaps the central – and inherently thorny – issue in sorting out

possessive expressions in any language is the question of how we define the

conceptual category to begin with. The present analysis is based on the notion of

possessive prototype understood as an experiential gestalt, which takes the PR to

be necessarily an animate (human) entity and the PM to be placed somewhere

along the possessibility hierarchy, without stating categorically what may or may

not count (universally) as a PM. Based on the corpus attestations of the Czech

patterns, we can draw at least two conclusions about possessibility. (i) Rather than

relying on purely linguistic categories, such as (in)animacy, concreteness, control,

etc., possessibility is best defined in terms of culturally determined clusters of

concepts and expectations about what is conventionally construed as possessible;

the conventional understanding can be then extended to cases where possessive

construal is conditioned by the type of discourse or genre. And (ii) different

grammatical forms expressing possession may interact with the possessibility

hierarchy in different ways.

These generalizations are based on a close study of two syntactic patterns

that both express possession as a time-stable and presupposed relation. The

analysis has established that the patterns are not equivalent semantically or

pragmatically and, therefore, cannot be treated simply as structural variants of a

single possessive schema. Each pattern encodes a distinct conceptualization of a

possessive relation, compatible with different communicative contexts: plain

possession is expressed by a Genitive PR (GP), which shares certain features with

non-possessive genitives; situated possession – an idiosyncratic combination of

possession and affectedness – is expressed by dative-marked Affected PR (AP),

which shares certain features with non-possessive dative-marked roles. It follows

that an adequate representation of the speakers‟ understanding of these patterns

29

requires reference to several layers of information: semantic and structural limits

on the PM (AP), affectedness of the PR (AP), semantic and morphological

constraints on the PR (GP), relative discourse prominence of the PR vis-à-vis the

PM (AP, GP), verb semantics (AP), and contextual compatibility (AP, GP).

The analysis thus makes an argument for taking a Construction Grammar

approach as a particularly useful way of framing our understanding of all the

relevant issues. First, the clusters of conventionally co-occuring features are

naturally captured through the notion of „grammatical construction‟. Second, the

two constructions constitute distinct pieces of a larger network of grammatical

entities organized around shared features, both formal and semantic: AP and GP

can be shown to occupy partially overlapping domains within the general

functional space of attributive possession. And finally, organizing our knowledge

about individual constructions in such a network provides us with a more refined

map of criteria that can play a role in expressing possessive relations in general.

Based on such a map, we can start articulating more systematic hypotheses about

the paths along which the prototype might be extended into more peripheral

instances; specifically, the extensions can be systematically motivated (and also

constrained) by the ways in which the constructions in the network may interact

both with each other and with the possessive prototype itself.

* I wish to thank all those colleagues whose questions, at various conference

gatherings, have helped me articulate my thoughts about this topic. And I‟m

especially grateful to Bill McGregor and Elizabeth Traugott for very careful

reading of the manuscript and helpful comments on the work.

Data:

Český národní korpus (spoken corpora PMK, BMK, ORAL2006; written corpora

SYN2000, SYN2006PUBL). Ústav Českého národního korpusu FF UK, Praha.

<http://ucnk.ff.cuni.cz>

Český národní korpus – A sample of Czech language and literature (SYNEK,

LITERA) [CD-ROM]. Ústav Českého národního korpusu FF UK, Praha.

<http://ucnk.ff.cuni.cz>

References:

Bally, Charles (1995/1926). The expression of concepts of the personal

domain and indivisibility in Indo-European languages. In Chappell, Hilary

& William B. McGregor (eds.), 31-61.

Berman, Ruth A. (1982). Dative marking of the affectee role: data from

Modern Hebrew. Hebrew Annual Review 6: 35-59.

Chappell, Hilary & William B. McGregor (eds.) (1995). The grammar of

inalienability: a typological perspective on body part terms and the part-

whole relation. Berlin: Mouton de Gruyter.

Croft, William (1985). Indirect Object „Lowering‟. Berkeley Linguistic Society

11: 39-51.

30

Croft, William (2001). Radical Construction Grammar. Syntactic theory in

typological perspective. Oxford: Oxford University Press.

Daneš, František (1974). Functional sentence perspective and the organization

of the text. In F. Daneš (ed.), Papers on functional sentence perspective,

106-128. Praha: Academia.

Fillmore, Charles J. (1982). Frame Semantics. In The Linguistic Society of

Korea (Ed.), Linguistics in the Morning Calm (111-137). Seoul: Hanshin.

Fillmore, Charles J. (1989). Grammatical Construction Theory and the familiar

dichotomies. In R. Dietrich & C.F. Graumann (Eds.), Language processing

in social context (17-38). Amsterdam: North-Holland/Elsevier.

Firbas, Jan (1966). On defining the theme in functional sentence perspective.

Travaux linguistiques de Prague 2: 267-280.

Fried, Mirjam (1994). BLS

Fried, Mirjam (1999a). From interest to ownership: a constructional view of

external PRs. In D.L. Payne & I. Barshi (Eds.), External possession (473-

504). Amsterdam & Philadelphia, PA: John Benjamins.

Fried, Mirjam (1999b). The „free‟ datives in Czech as a linking problem. In K.

Dziwirek, H. Coats, & C. Vakareliyska (Eds.), Annual Workshop on

Formal Approaches to Slavic Linguistics 7 (145-166). Ann Arbor:

Michigan Slavic Publications.

Fried, Mirjam (2008). Dative possessor: delimiting a grammatical category

based on usage. In F. Štícha & M. Fried (eds.), Gramatika a korpus /

Grammar and corpora (pp.). Praha: Academia.

Fried, Mirjam & Jan-Ola Östman (2004). Construction Grammar: a thumbnail

sketch. In M. Fried & J-O. Östman (eds.), Construction Grammar in a

cross-linguistic perspective. Amsterdam: John Benjamins, 11-86.

Grepl, Miroslav & Petr Karlík (1998). Skladba spisovné češtiny [Syntax of literary

Czech]. Praha: Votobia.

Heine, Bernd (1997). Possession: cognitive sources, forces, and

grammaticalization. Cambridge University Press.

Manoliu-Manea, Maria (1995). Inalienability and topicality in Romanian:

pragma-semantics of syntax. In Chappell, Hilary & William B. McGregor

(eds.), 711-743.

O‟Connor, Mary Catherine (1994). The marking of possession in Northern

Pomo: Privative opposition and pragmatic inference. BLS 20, 387-401.

Payne, Doris L. (1997a). The Maasai external PR construction. In J. Bybee, J.

Haiman & S.A. Thompson (Eds.), Essays on Language Function and

Language Type (395-422). Amsterdam & Philadelphia, PA: John

Benjamins.

Payne, Doris L. (1997b). Argument structure and locus of affect in the Maasai

External Possession Construction. BLS 23 [Special Session on Syntax and

Semantics in Africa], 98-115.

Payne, Doris L. & Immanuel Barshi (Eds.) (1999). External Possession.

Amsterdam & Philadelphia, PA: John Benjamins.

31

Pit‟ha, Petr (1992): Posesivní vztah v češtině. Praha: AVED.

Sgall, P., E. Hajičová & J. Panevová (1986). The meaning of the sentence in its

semantic and pragmatic aspects. D. Reidel Publishing Company.

Taylor, John R. (1989). Linguistic categorization: Prototypes in linguistic

theory. Oxford: Clarendon Press.

Tsunoda, Tasaku (1995). The possession cline in Japanese and other

languages. In Chappell, Hilary & W. McGregor (eds.), 565-630.

Zimek, Rudolf (1960): K chápání posesivnosti. Rusko-české studie, Sborník

Vysoké školy pedagogické v Praze, Jazyk a literatura II, 131-156.